[R-sig-ME] glmmTMB syntax to brm() syntax

[Ben Bolker]

   the way you would do this in glmmTMB (with simulate_new [sic], you 
wouldn't need a different function:

  * conditional model: ~condition + Cognitive_Rate ... (not surprisingly)
  * random effect model :  ... + (1|subject) (also not surprising
  * varIdent:  dispformula = ~  (1|condition)


The tricky part is the corSymm part.  To do this you'd probably need to 
add an 'observation number within subject' variable (say it's called 
`obs`) and use  ... + (0 + obs | subject).

   I don't know if that would correspond exactly. You could try some 
experiments.

   If not, you might have to implement the simulation code yourself ...

   Ben Bolker

On 10/26/24 21:51, Simon Harmel wrote:
> Ben, can I ask a similar question? Is there also a function similar to
> glmmTMB::new_simulate() to easily simulate the following model?
> 
> nlme::lme(Y ~ condition + Cognitive_Rate, random= ~1 | subject, data = DATA,
> weights = varIdent(form=~1|condition),
>   correlation = corSymm(form = ~1| subject))
> 
> Thank you!
> Simon
> 
> On Thu, Oct 24, 2024 at 9:11 PM Ben Bolker <bbolker using gmail.com> wrote:
>>
>>     Most of the time taken by the brms version is in compilation. It may
>> be possible (I don't remember how) to cache the compiled model and
>> re-use it for subsequent models, if you are going to be (for example)
>> fitting the same model to many different data sets ...
>>
>> On 10/24/24 21:44, Simon Harmel wrote:
>>> Thank you so very much, Ben! And wow, the brm() version is extremely slow.
>>>
>>> Simon
>>>
>>> On Thu, Oct 24, 2024 at 11:11 AM Ben Bolker <bbolker using gmail.com> wrote:
>>>>
>>>>      See below.  The two models (glmmTMB and brms) give sufficiently
>>>> similar estimates that I'm confident that the specifications match.
>>>>
>>>> set.seed(101)
>>>> library(glmmTMB)
>>>> library(brms)
>>>> library(broom.mixed)
>>>> library(tidyverse)
>>>>
>>>> dd <- data.frame(ID = rep(1:100, each = 10),
>>>>                     TRIAL_INDEX = rep(1:10, 100),
>>>>                     con = rnorm(1000))
>>>> dd$pic_percent <- simulate_new(
>>>>        ~ con + (0+con | ID) +
>>>>            (0+con | TRIAL_INDEX),
>>>>        ziformula = ~1,
>>>>        family = beta_family(),
>>>>        newdata = dd,
>>>>        newparams = list(beta = c(0, 0.5), theta = rep(-1,2),
>>>>                         betadisp = 1, betazi = -2))[[1]]
>>>>
>>>>
>>>> m1 <- glmmTMB(pic_percent ~ con + (0+con | ID) +
>>>>        (0+con | TRIAL_INDEX),
>>>>            data=dd,
>>>>            family = beta_family(),
>>>>            ziformula = ~1)
>>>>
>>>> ##
>>>> https://mvuorre.github.io/posts/2019-02-18-analyze-analog-scale-ratings-with-zero-one-inflated-beta-models/
>>>> m2 <- brm(
>>>>        bf(pic_percent ~ con + (0+con | ID) +
>>>>               (0+con | TRIAL_INDEX),
>>>>        zi = ~ 1),
>>>>        data=dd,
>>>>        family = zero_inflated_beta()
>>>> )
>>>>
>>>>
>>>> (purrr::map_dfr(list(glmmTMB = m1, brms = m2), tidy, .id = "model")
>>>>        |> select(model, effect, component, group, term, estimate)
>>>>        |> pivot_wider(names_from = model, values_from = estimate)
>>>> )
>>>>
>>>>
>>>>
>>>> On 10/23/24 19:13, Simon Harmel wrote:
>>>>> Hello all,
>>>>>
>>>>> I was wondering what is the closest equivalent of my glmmTMB syntax below
>>>>> in brms::brm() syntax?
>>>>>
>>>>> glmmTMBglmmTMB(pic_percent ~ con +
>>>>>                          (0+con | ID) +
>>>>>                          (0+con | TRIAL_INDEX),
>>>>>                        data=DATA,
>>>>>            family = beta_family(),
>>>>>            ziformula = ~1)
>>>>>
>>>>>
>>>>> Thank you,
>>>>>
>>>>> Simon
>>>>>
>>>>>         [[alternative HTML version deleted]]
>>>>>
>>>>> _______________________________________________
>>>>> R-sig-mixed-models using r-project.org mailing list
>>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>>>>
>>>> --
>>>> Dr. Benjamin Bolker
>>>> Professor, Mathematics & Statistics and Biology, McMaster University
>>>> Director, School of Computational Science and Engineering
>>>> * E-mail is sent at my convenience; I don't expect replies outside of
>>>> working hours.
>>>>
>>>> _______________________________________________
>>>> R-sig-mixed-models using r-project.org mailing list
>>>> https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
>>
>> --
>> Dr. Benjamin Bolker
>> Professor, Mathematics & Statistics and Biology, McMaster University
>> Director, School of Computational Science and Engineering
>> * E-mail is sent at my convenience; I don't expect replies outside of
>> working hours.
>>

-- 
Dr. Benjamin Bolker
Professor, Mathematics & Statistics and Biology, McMaster University
Director, School of Computational Science and Engineering
* E-mail is sent at my convenience; I don't expect replies outside of 
working hours.